Abstract: Academician Lu Ruqian proposed Full Cycle Computer Aided Animation Auto-generation Technology based on artificial intelligence technology. The technology needs to recognize special types of named entities such as school names, restaurant names and emporium names, which can be performed in animation. A method for recognizing special types of named entities is proposed based on the Hidden Markov Model（HMM） and rules. The method uses HMM and trains the model with part-of-speech information, feature words and the sense of words, then complements and corrects the recognition results of HMM with rules automatically extracted from the training corpus and entity names. The accuracy, recall and F-score of the open testing experiment reaches 79.89%, 86.6% and 83.11%.

Key words: Hidden Markov Model（HMM）, special kinds of named entity, expanded part-of-speech, animation auto-generation

摘要： 基于人工智能技术，陆汝钤院士提出了全过程计算机辅助动画自动生成技术。该技术需要对能在动画中具体表现的特殊类型命名实体进行识别，如学校名称、餐馆名称、商场名称等。提出了一种基于隐马尔科夫模型（HMM）和规则相结合的特殊类型命名实体识别方法，利用词性、特征词和词义等信息对HMM模型进行训练，并用自动提取的规则对统计模型的识别结果进行补充和修正。开放性测试实验的最高准确率、召回率和F值分别达到了79.89%、86.6%、83.11%。

关键词: 隐马尔科夫模型, 特殊实体, 扩充词性, 动画自动生成